AI Model Rankings
An independent leaderboard - models scored on a hybrid of public benchmarks, verified reviews and community votes. Transparent by design: every model links to its full profile.
| Vote | # | Model | Score | Capability | Rating | ARC-AGI-3 (Provider Adapter harness) | ARC-AGI-3 (Standard harness) | Artificial Analysis Intelligence Index | Artificial Analysis Intelligence Index v4.3.2 | ArXivMath | AutomationBench |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | 96 +33 vs avg | 96 | - | - | - | - | - | - | - | ||
| 2 | 89 +26 vs avg | 89 | - | - | - | - | - | - | - | ||
| 3 | 87 +23 vs avg | 87 | - | 99.9 % | 62.7 % | - | - | - | - | ||
| 4 | Shanghai Artificial Intelligence Laboratory · 744B total, mixture of experts, built on GLM-5.2 (active parameters not published) params · 256,000 ctx · MIT | 87 +23 vs avg | 87 | - | - | - | - | - | - | - | |
| 5 | Alibaba · 2.4T (95B active) params · 1,000,000 ctx · Proprietary (weights announced) | 86 +23 vs avg | 86 | - | - | - | - | - | - | - | |
| 6 | 85 +21 vs avg | 85 | - | - | - | - | - | - | - | ||
| 7 | 84 +20 vs avg | 80 | - | - | - | - | - | - | - | ||
| 8 | 80 +17 vs avg | 77 | - | - | - | - | - | - | - | ||
| 9 | 73 +9 vs avg | 73 | - | - | - | - | - | - | - | ||
| 10 | 71 +7 vs avg | 71 | - | - | - | - | - | - | - | ||
| 11 | 68 +5 vs avg | 68 | - | - | - | - | - | 88 % | - | ||
| 12 | 64 +0 vs avg | 64 | - | - | - | - | - | - | - | ||
| 13 | 64 +0 vs avg | 61 | - | - | - | - | - | - | - | ||
| 14 | 61 −3 vs avg | 61 | - | - | - | - | - | - | - | ||
| 15 | 58 −5 vs avg | 58 | - | - | - | - | 58 index +9.6 | - | - | ||
| 16 | 57 −6 vs avg | 57 | - | - | - | - | - | - | - | ||
| 17 | 56 −7 vs avg | 56 | - | - | - | - | 56 index +7.6 | - | - | ||
| 18 | 53 −10 vs avg | 53 | - | - | - | - | - | - | 31.4 % | ||
| 19 | 53 −10 vs avg | 53 | - | - | - | - | 53 index +4.6 | - | - | ||
| 20 | 53 −11 vs avg | 53 | - | - | - | 61 index | 44 index −4.4 | - | - | ||
| 21 | 52 −11 vs avg | 52 | - | - | - | - | 52 index +3.6 | - | - | ||
| 22 | 49 −14 vs avg | 61 | - | - | - | - | - | - | - | ||
| 23 | 48 −15 vs avg | 48 | - | - | - | - | 48 index −0.4 | - | - | ||
| 24 | 48 −16 vs avg | 48 | - | - | - | - | - | - | - | ||
| 25 | 47 −16 vs avg | 47 | - | - | - | - | - | - | - | ||
| 26 | 46 −17 vs avg | 46 | - | - | - | - | 46 index −2.4 | - | - | ||
| 27 | Xiaomi · 1.02T total, 42B active (sparse mixture of experts) params · 1,000,000 ctx · MIT — open weights (published as MiMo-V2.6-Pro-RL) | 46 −17 vs avg | 46 | - | - | - | - | 46 index −2.4 | - | - | |
| 28 | StepFun · 600B total, 27B active per token (sparse mixture of experts) params · 1,000,000 ctx · Proprietary — BF16 open weights announced for 15 October 2026, not published yet | 44 −19 vs avg | 44 | - | - | - | - | 44 index −4.4 | - | - | |
| 29 | 37 −26 vs avg | 37 | - | - | - | - | 37 index −11.4 | - | - | ||
| 30 | 0 | - | - | - | - | - | - | - | - | ||
| 31 | Agnes AI · 33B (open-weights Preview checkpoint; API model size not published) params · Apache 2.0 (Preview checkpoint weights) | 0 | - | - | - | - | - | - | - | - | |
| 32 | 0 | - | - | - | - | - | - | - | - | ||
| 33 | Cantina Security (with Yeta) · 321B total, 18B active (mixture of experts) params · 1,048,576 ctx · MIT | 0 | - | - | - | - | - | - | - | - | |
| 34 | 0 | - | - | - | - | - | - | - | - | ||
| 35 | PrismML · 27B, compressed to ternary weights: 1.76 effective bits per weight and a 5.9 GB footprint params · 262,144 ctx · Apache 2.0 | 0 | - | - | - | - | - | - | - | - | |
| 36 | 0 | - | - | - | - | - | - | - | - | ||
| 37 | 0 | - | - | - | - | - | - | - | - | ||
| 38 | 0 | - | - | - | - | - | - | - | - | ||
| 39 | 0 | - | - | - | - | - | - | - | - | ||
| 40 | 0 | - | - | - | - | - | - | - | - | ||
| 41 | Contrastive-LM · 8B (frozen Qwen3-8B encoder with two small trained heads) params · Apache-2.0 | 0 | - | - | - | - | - | - | - | - | |
| 42 | Cua · 706,048 trainable parameters — a 2.8 MB checkpoint params · MIT — weights, training code and dataset published | 0 | - | - | - | - | - | - | - | - | |
| 43 | 0 | - | - | - | - | - | - | - | - | ||
| 44 | 0 | - | - | - | - | - | - | - | - | ||
| 45 | VIDRAFT · 180B mixture of experts (10 of 512 routed experts active per token, plus a shared expert); a fine-tune of Qwen3.8-Flash-Next params · 262,144 ctx · Qwen Community License 1.0 | 0 | - | - | - | - | - | - | - | - | |
| 46 | 0 | - | - | - | - | - | - | - | - | ||
| 47 | 0 | - | - | - | - | - | - | - | - | ||
| 48 | 0 | - | - | - | - | - | - | - | - | ||
| 49 | 0 | - | - | - | - | - | - | - | - | ||
| 50 | 0 | - | - | - | - | - | - | - | - | ||
| 51 | 0 | - | - | - | - | - | - | - | - | ||
| 52 | 0 | - | - | - | - | - | - | - | - | ||
| 53 | 0 | - | - | - | - | - | - | - | - | ||
| 54 | 0 | - | - | - | - | - | - | - | - | ||
| 55 | 0 | - | - | - | - | - | - | - | - | ||
| 56 | 0 | - | - | - | - | - | - | - | - | ||
| 57 | 0 | - | - | - | - | - | - | - | - | ||
| 58 | 0 | - | - | - | - | - | - | - | - | ||
| 59 | 0 | - | - | - | - | - | - | - | - | ||
| 60 | 0 | - | - | - | - | - | - | - | - | ||
| 61 | 0 | - | - | - | - | - | - | - | - | ||
| 62 | 0 | - | - | - | - | - | - | - | - | ||
| 63 | Fastino · 340M (developer’s model card; the models page says 355M) params · Apache-2.0 | 0 | - | - | - | - | - | - | - | - | |
| 64 | 0 | - | - | - | - | - | - | - | - | ||
| 65 | Z.ai · 1,000,000 ctx · Proprietary — hosted tier, no weights of its own published | 0 | - | - | - | - | - | - | - | - | |
| 66 | 0 | - | - | - | - | - | - | - | - | ||
| 67 | 0 | - | - | - | - | - | - | - | - | ||
| 68 | IBM · Two sizes, built on the 8B and 30B Granite 4.1 base models, each with two activated-LoRA adapters and a 470M speech encoder params · 128,000 ctx · Apache 2.0 | 0 | - | - | - | - | - | - | - | - | |
| 69 | 0 | - | - | - | - | - | - | - | - | ||
| 70 | 0 | - | - | - | - | - | - | - | - | ||
| 71 | 0 | - | - | - | - | - | - | - | - | ||
| 72 | Tencent · 770B total, 49B active (mixture of experts) params · 1,000,000 ctx | 0 | - | - | - | - | - | - | - | - | |
| 73 | Bilibili (Index Team) · 35B total, 3B active (mixture of experts) params · Apache 2.0 | 0 | - | - | - | - | - | - | - | - | |
| 74 | InternLM · 0.8B, 2B and 4B (fine-tunes of Qwen3.5-0.8B, -2B and -4B) params · Apache-2.0 | 0 | - | - | - | - | - | - | - | - | |
| 75 | 0 | - | - | - | - | - | - | - | - | ||
| 76 | 0 | - | - | - | - | - | - | - | - | ||
| 77 | 0 | - | - | - | - | - | - | - | - | ||
| 78 | 0 | - | - | - | - | - | - | - | - | ||
| 79 | 0 | - | - | - | - | - | - | - | - | ||
| 80 | 0 | - | - | - | - | - | - | - | - | ||
| 81 | 0 | - | - | - | - | - | - | - | - | ||
| 82 | 0 | - | - | - | - | - | - | - | - | ||
| 83 | 0 | - | - | - | - | - | - | - | - | ||
| 84 | Liquid AI · 24B total parameters, about 2.45B active per token (mixture of experts, 64 experts per block, top-4 routing) params · 131,072 ctx · LFM 1.0 | 0 | - | - | - | - | - | - | - | - | |
| 85 | Light Origins · 5.5B (Hugging Face weight count; not stated by the developer) params · Apache 2.0 — open weights | 0 | - | - | - | - | - | - | - | - | |
| 86 | Paradigma · 1B (dense) params · 131,072 ctx · Apache 2.0 — open weights | 0 | - | - | - | - | - | - | - | - | |
| 87 | 0 | - | - | - | - | - | - | - | - | ||
| 88 | 0 | - | - | - | - | - | - | - | - | ||
| 89 | 0 | - | - | - | - | - | - | - | - | ||
| 90 | 0 | - | - | - | - | - | - | - | - | ||
| 91 | 0 | - | - | - | - | - | - | - | - | ||
| 92 | 0 | - | - | - | - | - | - | - | - | ||
| 93 | 0 | - | - | - | - | - | - | - | - | ||
| 94 | 0 | - | - | - | - | - | - | - | - | ||
| 95 | Xiaomi · 309B total, 15B active (sparse mixture of experts) params · 1,000,000 ctx · MIT — open weights (published as MiMo-V2.6-Flash-RL) | 0 | - | - | - | - | - | - | - | - | |
| 96 | 0 | - | - | - | - | - | - | - | - | ||
| 97 | 0 | - | - | - | - | - | - | - | - | ||
| 98 | 0 | - | - | - | - | - | - | - | - | ||
| 99 | 0 | - | - | - | - | - | - | - | - | ||
| 100 | Meta · 816B total parameters, 42B active per token (mixture of experts, 256 experts with 8 active per token) params · 1,000,000 ctx · Muse Model License | 0 | - | - | - | - | - | - | - | - | |
| 101 | Meta · 1,048,576 ctx · Proprietary (no licence stated; weights not published) | 0 | - | - | - | - | - | - | - | - | |
| 102 | NaiveAI · 309B total, 15.5B active (mixture of experts) params · 1,000,000 ctx · MIT | 0 | - | - | - | - | - | - | - | - | |
| 103 | 0 | - | - | - | - | - | - | - | - | ||
| 104 | 0 | - | - | - | - | - | - | - | - | ||
| 105 | 0 | - | - | - | - | - | - | - | - | ||
| 106 | NVIDIA · 120B total parameters, 12B active (hybrid Mamba-2, mixture of experts and attention layers, with a C-RADIO vision encoder) params · 1,000,000 ctx · OpenMDW-1.1 | 0 | - | - | - | - | - | - | - | - | |
| 107 | NVIDIA · 4B parameters; looped design with 8 input, 20 shared recurrent and 8 output layers params · 256,000 ctx · OpenMDW License Agreement 1.1 | 0 | - | - | - | - | - | - | - | - | |
| 108 | 0 | - | - | - | - | - | - | - | - | ||
| 109 | 0 | - | - | - | - | - | - | - | - | ||
| 110 | 0 | - | - | - | - | - | - | - | - | ||
| 111 | 0 | - | - | - | - | - | - | - | - | ||
| 112 | Alibaba · 7B in the visual generation component (32 single-stream DiT layers) params · Qwen Research Licence — non-commercial only; commercial use needs a separate licence | 0 | - | - | - | - | - | - | - | - | |
| 113 | 0 | - | - | - | - | - | - | - | - | ||
| 114 | 0 | - | - | - | - | - | - | - | - | ||
| 115 | 0 | - | - | - | - | - | - | - | - | ||
| 116 | 0 | - | - | - | - | - | - | - | - | ||
| 117 | 0 | - | - | - | - | - | - | - | - | ||
| 118 | 0 | - | - | - | - | - | - | - | - | ||
| 119 | 0 | - | - | - | - | - | - | - | - | ||
| 120 | 0 | - | - | - | - | - | - | - | - | ||
| 121 | 0 | - | - | - | - | - | - | - | - | ||
| 122 | BottleCap AI · 27.8B params · 262,144 ctx · PolyForm Small Business 1.0.0 | 0 | - | - | - | - | - | - | - | - | |
| 123 | 0 | - | - | - | - | - | - | - | - | ||
| 124 | Bad Theory Labs · 177B total, 6.6B active params · 262,144 ctx · Qwen Community License 1.0 — open weights | 0 | - | - | - | - | - | - | - | - | |
| 125 | 0 | - | - | - | - | - | - | - | - | ||
| 126 | 0 | - | - | - | - | - | - | - | - | ||
| 127 | Tencent Youtu Lab · 5B params · Youtu-Parsing licence (custom; not intended for use within the European Union) | 0 | - | - | - | - | - | - | - | - |
Score = hybrid of benchmark capability (60%), review rating (25%) and community votes (15%), renormalized by available signals. Capability is the mean of normalized benchmark results; green/red deltas compare each value to the average across benchmarked models. Official (editor-verified) and community-submitted values are kept separate - community submissions only affect rankings after editor approval. Your vote, rating and submitted stats all feed this score - sign-in required, so rankings stay hard to game.
Missing a model?
Suggest one and our editors will review and add it to the board.
New AI model launches
The models that launched on TrustList this month, voted up by the community.
- 1decision-machine-1AI model
CloudRaker's hosted decision model behind Milliseconds.ai, launched 19 September 2026, returning typed labels, yes/no answers, scores and extracted fields from text or images for $0.04 per million input tokens, with output free.
TL score - 2Step 5 PreviewAI model
A 600B mixture-of-experts flagship with 27B active per token and a million-token window, at $1.00 per million in and $2.70 out — priced well under the frontier it is measured against.
TL score - 3Qwen-Image-2.1AI model
A 7B image model that generates and edits in one checkpoint, with native transparent output — and, unlike every Qwen image release before it, a licence that forbids commercial use.
TL score - 4Qwen-Audio-3.1-Realtime-PlusAI model
Alibaba's full-duplex speech-to-speech model on Model Studio, released 20 September 2026, with a 262,144-token context, function calling, web search and voice cloning; audio output costs $24 per million tokens in Singapore.
TL score - 5Hemmingway-1AI model
A 27B open-weight fine-tune of Qwen3.8-27B for emails, messages and short stories, with a 262,144-token context; weights under CC BY-NC 4.0 since 22 September, hosted API $0.24 in and $0.90 out per million tokens.
TL score - 6KevAI model
Jared Palmer's open-source family of small decision models (0.8B, 4B and 9B) on Qwen3.5 that return typed yes/no, choice and score probabilities, released 20 September 2026 under Apache 2.0 with training code and evaluation data.
TL score - 7Grok 4.7AI model
The new Grok flagship, at Grok 4.6's price: $2 in and $6 out per million tokens under 200k-token prompts, a 500k context, reasoning from low to xhigh — and, on the Responses API, reasoning returned encrypted whether asked for or not.
TL score - 8Altar-1AI model
Aikido Security's open-weight security model, a pruned 4-bit version of GLM-5.3 that keeps 168 of 256 experts in 328 GB, released 21 September 2026 to run on-premises on four H200 GPUs under the GLM-5.3 licence.
TL score
More on TrustList
Everything here links back to the same verified catalogue. Pick your next stop.
- AI-model launchesEvery model launch, newest first.
- CompaniesAgencies, consultancies and IT service providers, ranked by verified reviews.
- ProductsSoftware and SaaS with pricing, features, integrations and alternatives.
- ArticlesGuides, comparisons and research from the editorial desk and the community.
- AwardsAnnual recognition decided by verified reviews and an independent jury.
- LaunchesNew products and releases, voted up by the community every day.
- RequestsBuyers describe what they need; vendors respond directly.
- PeopleReviewers, authors and makers with public profiles.
- ComparePut up to four listings side by side before you shortlist.